2 * 64-bit pSeries and RS/6000 setup code.
4 * Copyright (C) 1995 Linus Torvalds
5 * Adapted from 'alpha' version by Gary Thomas
6 * Modified by Cort Dougan (cort@cs.nmt.edu)
7 * Modified by PPC64 Team, IBM Corp
9 * This program is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU General Public License
11 * as published by the Free Software Foundation; either version
12 * 2 of the License, or (at your option) any later version.
16 * bootup setup stuff..
19 #include <linux/cpu.h>
20 #include <linux/errno.h>
21 #include <linux/sched.h>
22 #include <linux/kernel.h>
24 #include <linux/stddef.h>
25 #include <linux/unistd.h>
26 #include <linux/user.h>
27 #include <linux/tty.h>
28 #include <linux/major.h>
29 #include <linux/interrupt.h>
30 #include <linux/reboot.h>
31 #include <linux/init.h>
32 #include <linux/ioport.h>
33 #include <linux/console.h>
34 #include <linux/pci.h>
35 #include <linux/utsname.h>
36 #include <linux/adb.h>
37 #include <linux/export.h>
38 #include <linux/delay.h>
39 #include <linux/irq.h>
40 #include <linux/seq_file.h>
41 #include <linux/root_dev.h>
43 #include <linux/of_pci.h>
46 #include <asm/processor.h>
48 #include <asm/pgtable.h>
51 #include <asm/pci-bridge.h>
52 #include <asm/iommu.h>
54 #include <asm/machdep.h>
57 #include <asm/nvram.h>
61 #include <asm/ppc-pci.h>
62 #include <asm/i8259.h>
65 #include <asm/firmware.h>
68 #include <asm/plpar_wrappers.h>
69 #include <asm/kexec.h>
70 #include <asm/isa-bridge.h>
71 #include <asm/security_features.h>
77 unsigned long CMO_PageSize = (ASM_CONST(1) << IOMMU_PAGE_SHIFT_4K);
78 EXPORT_SYMBOL(CMO_PageSize);
80 int fwnmi_active; /* TRUE if an FWNMI handler is present */
82 static void pSeries_show_cpuinfo(struct seq_file *m)
84 struct device_node *root;
85 const char *model = "";
87 root = of_find_node_by_path("/");
89 model = of_get_property(root, "model", NULL);
90 seq_printf(m, "machine\t\t: CHRP %s\n", model);
93 seq_printf(m, "MMU\t\t: Radix\n");
95 seq_printf(m, "MMU\t\t: Hash\n");
98 /* Initialize firmware assisted non-maskable interrupts if
99 * the firmware supports this feature.
101 static void __init fwnmi_init(void)
103 unsigned long system_reset_addr, machine_check_addr;
105 int ibm_nmi_register = rtas_token("ibm,nmi-register");
106 if (ibm_nmi_register == RTAS_UNKNOWN_SERVICE)
109 /* If the kernel's not linked at zero we point the firmware at low
110 * addresses anyway, and use a trampoline to get to the real code. */
111 system_reset_addr = __pa(system_reset_fwnmi) - PHYSICAL_START;
112 machine_check_addr = __pa(machine_check_fwnmi) - PHYSICAL_START;
114 if (0 == rtas_call(ibm_nmi_register, 2, 1, NULL, system_reset_addr,
119 static void pseries_8259_cascade(struct irq_desc *desc)
121 struct irq_chip *chip = irq_desc_get_chip(desc);
122 unsigned int cascade_irq = i8259_irq();
125 generic_handle_irq(cascade_irq);
127 chip->irq_eoi(&desc->irq_data);
130 static void __init pseries_setup_i8259_cascade(void)
132 struct device_node *np, *old, *found = NULL;
133 unsigned int cascade;
135 unsigned long intack = 0;
138 for_each_node_by_type(np, "interrupt-controller") {
139 if (of_device_is_compatible(np, "chrp,iic")) {
146 printk(KERN_DEBUG "pic: no ISA interrupt controller\n");
150 cascade = irq_of_parse_and_map(found, 0);
152 printk(KERN_ERR "pic: failed to map cascade interrupt");
155 pr_debug("pic: cascade mapped to irq %d\n", cascade);
157 for (old = of_node_get(found); old != NULL ; old = np) {
158 np = of_get_parent(old);
162 if (strcmp(np->name, "pci") != 0)
164 addrp = of_get_property(np, "8259-interrupt-acknowledge", NULL);
167 naddr = of_n_addr_cells(np);
168 intack = addrp[naddr-1];
170 intack |= ((unsigned long)addrp[naddr-2]) << 32;
173 printk(KERN_DEBUG "pic: PCI 8259 intack at 0x%016lx\n", intack);
174 i8259_init(found, intack);
176 irq_set_chained_handler(cascade, pseries_8259_cascade);
179 static void __init pseries_init_irq(void)
181 /* Try using a XIVE if available, otherwise use a XICS */
182 if (!xive_spapr_init()) {
184 pseries_setup_i8259_cascade();
188 static void pseries_lpar_enable_pmcs(void)
190 unsigned long set, reset;
194 plpar_hcall_norets(H_PERFMON, set, reset);
197 static int pci_dn_reconfig_notifier(struct notifier_block *nb, unsigned long action, void *data)
199 struct of_reconfig_data *rd = data;
200 struct device_node *parent, *np = rd->dn;
205 case OF_RECONFIG_ATTACH_NODE:
206 parent = of_get_parent(np);
207 pdn = parent ? PCI_DN(parent) : NULL;
209 pci_add_device_node_info(pdn->phb, np);
213 case OF_RECONFIG_DETACH_NODE:
216 list_del(&pdn->list);
225 static struct notifier_block pci_dn_reconfig_nb = {
226 .notifier_call = pci_dn_reconfig_notifier,
229 struct kmem_cache *dtl_cache;
231 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
233 * Allocate space for the dispatch trace log for all possible cpus
234 * and register the buffers with the hypervisor. This is used for
235 * computing time stolen by the hypervisor.
237 static int alloc_dispatch_logs(void)
240 struct paca_struct *pp;
241 struct dtl_entry *dtl;
243 if (!firmware_has_feature(FW_FEATURE_SPLPAR))
249 for_each_possible_cpu(cpu) {
251 dtl = kmem_cache_alloc(dtl_cache, GFP_KERNEL);
253 pr_warn("Failed to allocate dispatch trace log for cpu %d\n",
255 pr_warn("Stolen time statistics will be unreliable\n");
260 pp->dispatch_log = dtl;
261 pp->dispatch_log_end = dtl + N_DISPATCH_LOG;
265 /* Register the DTL for the current (boot) cpu */
266 dtl = get_paca()->dispatch_log;
267 get_paca()->dtl_ridx = 0;
268 get_paca()->dtl_curr = dtl;
269 get_paca()->lppaca_ptr->dtl_idx = 0;
271 /* hypervisor reads buffer length from this field */
272 dtl->enqueue_to_dispatch_time = cpu_to_be32(DISPATCH_LOG_BYTES);
273 ret = register_dtl(hard_smp_processor_id(), __pa(dtl));
275 pr_err("WARNING: DTL registration of cpu %d (hw %d) failed "
276 "with %d\n", smp_processor_id(),
277 hard_smp_processor_id(), ret);
278 get_paca()->lppaca_ptr->dtl_enable_mask = 2;
282 #else /* !CONFIG_VIRT_CPU_ACCOUNTING_NATIVE */
283 static inline int alloc_dispatch_logs(void)
287 #endif /* CONFIG_VIRT_CPU_ACCOUNTING_NATIVE */
289 static int alloc_dispatch_log_kmem_cache(void)
291 dtl_cache = kmem_cache_create("dtl", DISPATCH_LOG_BYTES,
292 DISPATCH_LOG_BYTES, 0, NULL);
294 pr_warn("Failed to create dispatch trace log buffer cache\n");
295 pr_warn("Stolen time statistics will be unreliable\n");
299 return alloc_dispatch_logs();
301 machine_early_initcall(pseries, alloc_dispatch_log_kmem_cache);
303 static void pseries_lpar_idle(void)
306 * Default handler to go into low thread priority and possibly
307 * low power mode by ceding processor to hypervisor
310 /* Indicate to hypervisor that we are idle. */
311 get_lppaca()->idle = 1;
314 * Yield the processor to the hypervisor. We return if
315 * an external interrupt occurs (which are driven prior
316 * to returning here) or if a prod occurs from another
317 * processor. When returning here, external interrupts
322 get_lppaca()->idle = 0;
326 * Enable relocation on during exceptions. This has partition wide scope and
327 * may take a while to complete, if it takes longer than one second we will
328 * just give up rather than wasting any more time on this - if that turns out
329 * to ever be a problem in practice we can move this into a kernel thread to
330 * finish off the process later in boot.
332 void pseries_enable_reloc_on_exc(void)
335 unsigned int delay, total_delay = 0;
338 rc = enable_reloc_on_exceptions();
339 if (!H_IS_LONG_BUSY(rc)) {
341 pr_info("Relocation on exceptions not"
343 } else if (rc != H_SUCCESS) {
344 pr_warn("Unable to enable relocation"
345 " on exceptions: %ld\n", rc);
350 delay = get_longbusy_msecs(rc);
351 total_delay += delay;
352 if (total_delay > 1000) {
353 pr_warn("Warning: Giving up waiting to enable "
354 "relocation on exceptions (%u msec)!\n",
362 EXPORT_SYMBOL(pseries_enable_reloc_on_exc);
364 void pseries_disable_reloc_on_exc(void)
369 rc = disable_reloc_on_exceptions();
370 if (!H_IS_LONG_BUSY(rc))
372 mdelay(get_longbusy_msecs(rc));
375 pr_warn("Warning: Failed to disable relocation on exceptions: %ld\n",
378 EXPORT_SYMBOL(pseries_disable_reloc_on_exc);
380 #ifdef CONFIG_KEXEC_CORE
381 static void pSeries_machine_kexec(struct kimage *image)
383 if (firmware_has_feature(FW_FEATURE_SET_MODE))
384 pseries_disable_reloc_on_exc();
386 default_machine_kexec(image);
390 #ifdef __LITTLE_ENDIAN__
391 void pseries_big_endian_exceptions(void)
396 rc = enable_big_endian_exceptions();
397 if (!H_IS_LONG_BUSY(rc))
399 mdelay(get_longbusy_msecs(rc));
403 * At this point it is unlikely panic() will get anything
404 * out to the user, since this is called very late in kexec
405 * but at least this will stop us from continuing on further
406 * and creating an even more difficult to debug situation.
408 * There is a known problem when kdump'ing, if cpus are offline
409 * the above call will fail. Rather than panicking again, keep
410 * going and hope the kdump kernel is also little endian, which
413 if (rc && !kdump_in_progress())
414 panic("Could not enable big endian exceptions");
417 void pseries_little_endian_exceptions(void)
422 rc = enable_little_endian_exceptions();
423 if (!H_IS_LONG_BUSY(rc))
425 mdelay(get_longbusy_msecs(rc));
428 ppc_md.progress("H_SET_MODE LE exception fail", 0);
429 panic("Could not enable little endian exceptions");
434 static void __init find_and_init_phbs(void)
436 struct device_node *node;
437 struct pci_controller *phb;
438 struct device_node *root = of_find_node_by_path("/");
440 for_each_child_of_node(root, node) {
441 if (node->type == NULL || (strcmp(node->type, "pci") != 0 &&
442 strcmp(node->type, "pciex") != 0))
445 phb = pcibios_alloc_controller(node);
449 pci_process_bridge_OF_ranges(phb, node, 0);
450 isa_bridge_find_early(phb);
451 phb->controller_ops = pseries_pci_controller_ops;
457 * PCI_PROBE_ONLY and PCI_REASSIGN_ALL_BUS can be set via properties
460 of_pci_check_probe_only();
463 static void init_cpu_char_feature_flags(struct h_cpu_char_result *result)
466 * The features below are disabled by default, so we instead look to see
467 * if firmware has *enabled* them, and set them if so.
469 if (result->character & H_CPU_CHAR_SPEC_BAR_ORI31)
470 security_ftr_set(SEC_FTR_SPEC_BAR_ORI31);
472 if (result->character & H_CPU_CHAR_BCCTRL_SERIALISED)
473 security_ftr_set(SEC_FTR_BCCTRL_SERIALISED);
475 if (result->character & H_CPU_CHAR_L1D_FLUSH_ORI30)
476 security_ftr_set(SEC_FTR_L1D_FLUSH_ORI30);
478 if (result->character & H_CPU_CHAR_L1D_FLUSH_TRIG2)
479 security_ftr_set(SEC_FTR_L1D_FLUSH_TRIG2);
481 if (result->character & H_CPU_CHAR_L1D_THREAD_PRIV)
482 security_ftr_set(SEC_FTR_L1D_THREAD_PRIV);
484 if (result->character & H_CPU_CHAR_COUNT_CACHE_DISABLED)
485 security_ftr_set(SEC_FTR_COUNT_CACHE_DISABLED);
488 * The features below are enabled by default, so we instead look to see
489 * if firmware has *disabled* them, and clear them if so.
491 if (!(result->behaviour & H_CPU_BEHAV_FAVOUR_SECURITY))
492 security_ftr_clear(SEC_FTR_FAVOUR_SECURITY);
494 if (!(result->behaviour & H_CPU_BEHAV_L1D_FLUSH_PR))
495 security_ftr_clear(SEC_FTR_L1D_FLUSH_PR);
497 if (!(result->behaviour & H_CPU_BEHAV_BNDS_CHK_SPEC_BAR))
498 security_ftr_clear(SEC_FTR_BNDS_CHK_SPEC_BAR);
501 void pseries_setup_rfi_flush(void)
503 struct h_cpu_char_result result;
504 enum l1d_flush_type types;
509 * Set features to the defaults assumed by init_cpu_char_feature_flags()
510 * so it can set/clear again any features that might have changed after
511 * migration, and in case the hypercall fails and it is not even called.
513 powerpc_security_features = SEC_FTR_DEFAULT;
515 rc = plpar_get_cpu_characteristics(&result);
517 init_cpu_char_feature_flags(&result);
520 * We're the guest so this doesn't apply to us, clear it to simplify
521 * handling of it elsewhere.
523 security_ftr_clear(SEC_FTR_L1D_FLUSH_HV);
525 types = L1D_FLUSH_FALLBACK;
527 if (security_ftr_enabled(SEC_FTR_L1D_FLUSH_TRIG2))
528 types |= L1D_FLUSH_MTTRIG;
530 if (security_ftr_enabled(SEC_FTR_L1D_FLUSH_ORI30))
531 types |= L1D_FLUSH_ORI;
533 enable = security_ftr_enabled(SEC_FTR_FAVOUR_SECURITY) && \
534 security_ftr_enabled(SEC_FTR_L1D_FLUSH_PR);
536 setup_rfi_flush(types, enable);
539 #ifdef CONFIG_PCI_IOV
540 enum rtas_iov_fw_value_map {
541 NUM_RES_PROPERTY = 0, /* Number of Resources */
542 LOW_INT = 1, /* Lowest 32 bits of Address */
543 START_OF_ENTRIES = 2, /* Always start of entry */
544 APERTURE_PROPERTY = 2, /* Start of entry+ to Aperture Size */
545 WDW_SIZE_PROPERTY = 4, /* Start of entry+ to Window Size */
546 NEXT_ENTRY = 7 /* Go to next entry on array */
549 enum get_iov_fw_value_index {
550 BAR_ADDRS = 1, /* Get Bar Address */
551 APERTURE_SIZE = 2, /* Get Aperture Size */
552 WDW_SIZE = 3 /* Get Window Size */
555 resource_size_t pseries_get_iov_fw_value(struct pci_dev *dev, int resno,
556 enum get_iov_fw_value_index value)
559 struct device_node *dn = pci_device_to_OF_node(dev);
560 int i, num_res, ret = 0;
562 indexes = of_get_property(dn, "ibm,open-sriov-vf-bar-info", NULL);
567 * First element in the array is the number of Bars
568 * returned. Search through the list to find the matching
571 num_res = of_read_number(&indexes[NUM_RES_PROPERTY], 1);
572 if (resno >= num_res)
573 return 0; /* or an errror */
575 i = START_OF_ENTRIES + NEXT_ENTRY * resno;
578 ret = of_read_number(&indexes[i], 2);
581 ret = of_read_number(&indexes[i + APERTURE_PROPERTY], 2);
584 ret = of_read_number(&indexes[i + WDW_SIZE_PROPERTY], 2);
591 void of_pci_set_vf_bar_size(struct pci_dev *dev, const int *indexes)
593 struct resource *res;
594 resource_size_t base, size;
597 num_res = of_read_number(&indexes[NUM_RES_PROPERTY], 1);
598 num_res = min_t(int, num_res, PCI_SRIOV_NUM_BARS);
599 for (i = START_OF_ENTRIES, r = 0; r < num_res && r < PCI_SRIOV_NUM_BARS;
600 i += NEXT_ENTRY, r++) {
601 res = &dev->resource[r + PCI_IOV_RESOURCES];
602 base = of_read_number(&indexes[i], 2);
603 size = of_read_number(&indexes[i + APERTURE_PROPERTY], 2);
604 res->flags = pci_parse_of_flags(of_read_number
605 (&indexes[i + LOW_INT], 1), 0);
606 res->flags |= (IORESOURCE_MEM_64 | IORESOURCE_PCI_FIXED);
607 res->name = pci_name(dev);
609 res->end = base + size - 1;
613 void of_pci_parse_iov_addrs(struct pci_dev *dev, const int *indexes)
615 struct resource *res, *root, *conflict;
616 resource_size_t base, size;
620 * First element in the array is the number of Bars
621 * returned. Search through the list to find the matching
622 * bars assign them from firmware into resources structure.
624 num_res = of_read_number(&indexes[NUM_RES_PROPERTY], 1);
625 for (i = START_OF_ENTRIES, r = 0; r < num_res && r < PCI_SRIOV_NUM_BARS;
626 i += NEXT_ENTRY, r++) {
627 res = &dev->resource[r + PCI_IOV_RESOURCES];
628 base = of_read_number(&indexes[i], 2);
629 size = of_read_number(&indexes[i + WDW_SIZE_PROPERTY], 2);
630 res->name = pci_name(dev);
632 res->end = base + size - 1;
633 root = &iomem_resource;
635 "pSeries IOV BAR %d: trying firmware assignment %pR\n",
636 r + PCI_IOV_RESOURCES, res);
637 conflict = request_resource_conflict(root, res);
640 "BAR %d: %pR conflicts with %s %pR\n",
641 r + PCI_IOV_RESOURCES, res,
642 conflict->name, conflict);
643 res->flags |= IORESOURCE_UNSET;
648 static void pseries_pci_fixup_resources(struct pci_dev *pdev)
651 struct device_node *dn = pci_device_to_OF_node(pdev);
653 /*Firmware must support open sriov otherwise dont configure*/
654 indexes = of_get_property(dn, "ibm,open-sriov-vf-bar-info", NULL);
657 /* Assign the addresses from device tree*/
658 of_pci_set_vf_bar_size(pdev, indexes);
661 static void pseries_pci_fixup_iov_resources(struct pci_dev *pdev)
664 struct device_node *dn = pci_device_to_OF_node(pdev);
666 if (!pdev->is_physfn || pdev->is_added)
668 /*Firmware must support open sriov otherwise dont configure*/
669 indexes = of_get_property(dn, "ibm,open-sriov-vf-bar-info", NULL);
672 /* Assign the addresses from device tree*/
673 of_pci_parse_iov_addrs(pdev, indexes);
676 static resource_size_t pseries_pci_iov_resource_alignment(struct pci_dev *pdev,
680 struct device_node *dn = pci_device_to_OF_node(pdev);
682 /*Firmware must support open sriov otherwise report regular alignment*/
683 reg = of_get_property(dn, "ibm,is-open-sriov-pf", NULL);
685 return pci_iov_resource_size(pdev, resno);
687 if (!pdev->is_physfn)
689 return pseries_get_iov_fw_value(pdev,
690 resno - PCI_IOV_RESOURCES,
695 static void __init pSeries_setup_arch(void)
697 set_arch_panic_timeout(10, ARCH_PANIC_TIMEOUT);
699 /* Discover PIC type and setup ppc_md accordingly */
703 /* openpic global configuration register (64-bit format). */
704 /* openpic Interrupt Source Unit pointer (64-bit format). */
705 /* python0 facility area (mmio) (64-bit format) REAL address. */
707 /* init to some ~sane value until calibrate_delay() runs */
708 loops_per_jiffy = 50000000;
712 pseries_setup_rfi_flush();
714 /* By default, only probe PCI (can be overridden by rtas_pci) */
715 pci_add_flags(PCI_PROBE_ONLY);
717 /* Find and initialize PCI host bridges */
718 init_pci_config_tokens();
719 find_and_init_phbs();
720 of_reconfig_notifier_register(&pci_dn_reconfig_nb);
722 pSeries_nvram_init();
724 if (firmware_has_feature(FW_FEATURE_LPAR)) {
725 vpa_init(boot_cpuid);
726 ppc_md.power_save = pseries_lpar_idle;
727 ppc_md.enable_pmcs = pseries_lpar_enable_pmcs;
728 #ifdef CONFIG_PCI_IOV
729 ppc_md.pcibios_fixup_resources =
730 pseries_pci_fixup_resources;
731 ppc_md.pcibios_fixup_sriov =
732 pseries_pci_fixup_iov_resources;
733 ppc_md.pcibios_iov_resource_alignment =
734 pseries_pci_iov_resource_alignment;
737 /* No special idle routine */
738 ppc_md.enable_pmcs = power4_enable_pmcs;
741 ppc_md.pcibios_root_bridge_prepare = pseries_root_bridge_prepare;
744 static void pseries_panic(char *str)
746 panic_flush_kmsg_end();
750 static int __init pSeries_init_panel(void)
752 /* Manually leave the kernel version on the panel. */
753 #ifdef __BIG_ENDIAN__
754 ppc_md.progress("Linux ppc64\n", 0);
756 ppc_md.progress("Linux ppc64le\n", 0);
758 ppc_md.progress(init_utsname()->version, 0);
762 machine_arch_initcall(pseries, pSeries_init_panel);
764 static int pseries_set_dabr(unsigned long dabr, unsigned long dabrx)
766 return plpar_hcall_norets(H_SET_DABR, dabr);
769 static int pseries_set_xdabr(unsigned long dabr, unsigned long dabrx)
771 /* Have to set at least one bit in the DABRX according to PAPR */
772 if (dabrx == 0 && dabr == 0)
774 /* PAPR says we can only set kernel and user bits */
775 dabrx &= DABRX_KERNEL | DABRX_USER;
777 return plpar_hcall_norets(H_SET_XDABR, dabr, dabrx);
780 static int pseries_set_dawr(unsigned long dawr, unsigned long dawrx)
782 /* PAPR says we can't set HYP */
785 return plpar_set_watchpoint0(dawr, dawrx);
788 #define CMO_CHARACTERISTICS_TOKEN 44
789 #define CMO_MAXLENGTH 1026
791 void pSeries_coalesce_init(void)
793 struct hvcall_mpp_x_data mpp_x_data;
795 if (firmware_has_feature(FW_FEATURE_CMO) && !h_get_mpp_x(&mpp_x_data))
796 powerpc_firmware_features |= FW_FEATURE_XCMO;
798 powerpc_firmware_features &= ~FW_FEATURE_XCMO;
802 * fw_cmo_feature_init - FW_FEATURE_CMO is not stored in ibm,hypertas-functions,
803 * handle that here. (Stolen from parse_system_parameter_string)
805 static void pSeries_cmo_feature_init(void)
807 char *ptr, *key, *value, *end;
809 int page_order = IOMMU_PAGE_SHIFT_4K;
811 pr_debug(" -> fw_cmo_feature_init()\n");
812 spin_lock(&rtas_data_buf_lock);
813 memset(rtas_data_buf, 0, RTAS_DATA_BUF_SIZE);
814 call_status = rtas_call(rtas_token("ibm,get-system-parameter"), 3, 1,
816 CMO_CHARACTERISTICS_TOKEN,
820 if (call_status != 0) {
821 spin_unlock(&rtas_data_buf_lock);
822 pr_debug("CMO not available\n");
823 pr_debug(" <- fw_cmo_feature_init()\n");
827 end = rtas_data_buf + CMO_MAXLENGTH - 2;
828 ptr = rtas_data_buf + 2; /* step over strlen value */
831 while (*ptr && (ptr <= end)) {
832 /* Separate the key and value by replacing '=' with '\0' and
833 * point the value at the string after the '='
838 } else if (ptr[0] == '\0' || ptr[0] == ',') {
839 /* Terminate the string containing the key/value pair */
843 pr_debug("Malformed key/value pair\n");
844 /* Never found a '=', end processing */
848 if (0 == strcmp(key, "CMOPageSize"))
849 page_order = simple_strtol(value, NULL, 10);
850 else if (0 == strcmp(key, "PrPSP"))
851 CMO_PrPSP = simple_strtol(value, NULL, 10);
852 else if (0 == strcmp(key, "SecPSP"))
853 CMO_SecPSP = simple_strtol(value, NULL, 10);
854 value = key = ptr + 1;
859 /* Page size is returned as the power of 2 of the page size,
860 * convert to the page size in bytes before returning
862 CMO_PageSize = 1 << page_order;
863 pr_debug("CMO_PageSize = %lu\n", CMO_PageSize);
865 if (CMO_PrPSP != -1 || CMO_SecPSP != -1) {
866 pr_info("CMO enabled\n");
867 pr_debug("CMO enabled, PrPSP=%d, SecPSP=%d\n", CMO_PrPSP,
869 powerpc_firmware_features |= FW_FEATURE_CMO;
870 pSeries_coalesce_init();
872 pr_debug("CMO not enabled, PrPSP=%d, SecPSP=%d\n", CMO_PrPSP,
874 spin_unlock(&rtas_data_buf_lock);
875 pr_debug(" <- fw_cmo_feature_init()\n");
879 * Early initialization. Relocation is on but do not reference unbolted pages
881 static void __init pseries_init(void)
883 pr_debug(" -> pseries_init()\n");
885 #ifdef CONFIG_HVC_CONSOLE
886 if (firmware_has_feature(FW_FEATURE_LPAR))
887 hvc_vio_init_early();
889 if (firmware_has_feature(FW_FEATURE_XDABR))
890 ppc_md.set_dabr = pseries_set_xdabr;
891 else if (firmware_has_feature(FW_FEATURE_DABR))
892 ppc_md.set_dabr = pseries_set_dabr;
894 if (firmware_has_feature(FW_FEATURE_SET_MODE))
895 ppc_md.set_dawr = pseries_set_dawr;
897 pSeries_cmo_feature_init();
898 iommu_init_early_pSeries();
900 pr_debug(" <- pseries_init()\n");
904 * pseries_power_off - tell firmware about how to power off the system.
906 * This function calls either the power-off rtas token in normal cases
907 * or the ibm,power-off-ups token (if present & requested) in case of
908 * a power failure. If power-off token is used, power on will only be
909 * possible with power button press. If ibm,power-off-ups token is used
910 * it will allow auto poweron after power is restored.
912 static void pseries_power_off(void)
915 int rtas_poweroff_ups_token = rtas_token("ibm,power-off-ups");
917 if (rtas_flash_term_hook)
918 rtas_flash_term_hook(SYS_POWER_OFF);
920 if (rtas_poweron_auto == 0 ||
921 rtas_poweroff_ups_token == RTAS_UNKNOWN_SERVICE) {
922 rc = rtas_call(rtas_token("power-off"), 2, 1, NULL, -1, -1);
923 printk(KERN_INFO "RTAS power-off returned %d\n", rc);
925 rc = rtas_call(rtas_poweroff_ups_token, 0, 1, NULL);
926 printk(KERN_INFO "RTAS ibm,power-off-ups returned %d\n", rc);
931 static int __init pSeries_probe(void)
933 const char *dtype = of_get_property(of_root, "device_type", NULL);
937 if (strcmp(dtype, "chrp"))
940 /* Cell blades firmware claims to be chrp while it's not. Until this
941 * is fixed, we need to avoid those here.
943 if (of_machine_is_compatible("IBM,CPBW-1.0") ||
944 of_machine_is_compatible("IBM,CBEA"))
947 pm_power_off = pseries_power_off;
949 pr_debug("Machine is%s LPAR !\n",
950 (powerpc_firmware_features & FW_FEATURE_LPAR) ? "" : " not");
957 static int pSeries_pci_probe_mode(struct pci_bus *bus)
959 if (firmware_has_feature(FW_FEATURE_LPAR))
960 return PCI_PROBE_DEVTREE;
961 return PCI_PROBE_NORMAL;
964 struct pci_controller_ops pseries_pci_controller_ops = {
965 .probe_mode = pSeries_pci_probe_mode,
968 define_machine(pseries) {
970 .probe = pSeries_probe,
971 .setup_arch = pSeries_setup_arch,
972 .init_IRQ = pseries_init_irq,
973 .show_cpuinfo = pSeries_show_cpuinfo,
974 .log_error = pSeries_log_error,
975 .pcibios_fixup = pSeries_final_fixup,
976 .restart = rtas_restart,
978 .panic = pseries_panic,
979 .get_boot_time = rtas_get_boot_time,
980 .get_rtc_time = rtas_get_rtc_time,
981 .set_rtc_time = rtas_set_rtc_time,
982 .calibrate_decr = generic_calibrate_decr,
983 .progress = rtas_progress,
984 .system_reset_exception = pSeries_system_reset_exception,
985 .machine_check_exception = pSeries_machine_check_exception,
986 #ifdef CONFIG_KEXEC_CORE
987 .machine_kexec = pSeries_machine_kexec,
988 .kexec_cpu_down = pseries_kexec_cpu_down,
990 #ifdef CONFIG_MEMORY_HOTPLUG_SPARSE
991 .memory_block_size = pseries_memory_block_size,